Download Average power consumption breakdown of Wireless Sensor

Average power consumption breakdown of Wireless
Sensor Network nodes using IPv6 over LLNs
Javier Schandy
Leonardo Steinfeld
Fernando Silveira
Facultad de Ingenierı́a
Instituto de Ingenierı́a Eléctrica
Universidad de la República
Montevideo, Uruguay
Email: [email protected]
Facultad de Ingenierı́a
Instituto de Ingenierı́a Eléctrica
Universidad de la República
Montevideo, Uruguay
Email: [email protected]
Facultad de Ingenierı́a
Instituto de Ingenierı́a Eléctrica
Universidad de la República
Montevideo, Uruguay
Email: [email protected]
Abstract—Achieving low levels of energy consumption is
crucial for Wireless Sensor Networks in order to extend battery
lifetime. As network protocols become standardized, it becomes
mandatory to correctly select the optimal parameters in order to
reduce the energy consumption for a certain application.
This paper presents a simple but still powerful approach
for the analysis of the average power consumption of a sensor
node using the IPv6 over Low power and Lossy Networks (LLN)
stack, which is one of the most widely adopted and promising
communication stacks. Power consumption is broken down according to the node states (i.e. CPU, IRQ, LPM, Tx, Rx) and
according to the network protocols (e.g. CoAP, RPL, 6LoWPAN,
ContikiMAC), identifying the relative weight of each protocol in
the total energy consumption for several configurations. Results
show that the Low Power Listening (LPL) mechanism of the radio
duty cycling layer and RPL control messages have the highest
impact on the total energy consumption, while the application’s
report rate has a very low impact for periods over 60 seconds.
I.
I NTRODUCTION
Wireless sensor networks are becoming a very significant
enabling technology in many sectors but they present many
challenges due to some of their inherent characteristics. The
nodes comprising a wireless sensor network are expected
to be small, reliable, low cost, and low power. Since the
nodes are usually placed in hard-to-reach areas and in large
number, they are powered from batteries or harvest energy
from the environment [1]. In both cases reduced levels of
power consumption is mandatory, in the former case to avoid
frequent batteries or nodes replacement, and in latter case to
suit the scarce available energy or to minimize the harvesting
system cost and size (e.g. solar panel).
A basic sensor node is composed of the following building
blocks: a processing element with radio (RF transceiver), sensors/actuators and a power supply subsystem [2]. Node power
consumption results from the sum of the power contributions of
its electronic components, which in turn depend on the component state and the actual operation performed. The power
profile drain, i.e. the instantaneous power as a function of the
time, determines the effective node energy consumption. The
hardware together with software and external factors, such as
the environment and the interaction with the network, dictates
the power profile drain of each node. The hardware defines
the power consumption for each operating mode. The network
design and the communication protocols influence the behavior
of the nodes. The message exchange within or between the
nodes determine the state and actions of the node, particularly
the radio operation mode, i.e. receive, transmit or sleep mode,
and the microcontroller operation mode. Finally, the application is implemented on top of an operating system which
provides services to ease application development, including a
communication stack that implements network protocols. The
number of user messages injected on the network depends on
the application requirements.
Depending on the network characteristics and the application bandwidth requirements, a set of configuration parameters
can be chosen trading-off energy consumption with other
attributes like latency. Kovasch et al. [3] included the energy
consumption in their analysis of the CoAP protocol, but their
focus was in the system performance from an application
layer perspective and not from the whole network stack. Ali
[4] studies the RPL performance including several metrics;
however, their energy consumption analysis only considers the
time the radio is on without distinguishing between Reception
(Rx) and Transmission (Tx) states.
The main contribution of this paper is a deep study of
the energy breakdown of one of the most widely adopted
communication stacks in LLNs, which is shown on Table I.
The effect of varying some parameters, such as ContikiMAC’s
channel check rate (CCR) and the period of an observable
CoAP resource was also studied. The simulations’ results
were analyzed in order to break down the average power
consumption of the node’s radio in Tx and Rx states. The
results of the simulations show that modifying the default
Contiki1 CCR from 8Hz to 2Hz can achieve energy savings
of 30% in the total average power consumption in a typical
application. The effect of the application workload in energy
consumption was also studied, showing that for an observe
period of 60s, the application contribution is almost negligible,
suggesting that lower sample rates would not significantly
enhance the network lifetime.
The rest of the paper is organized as follows. In Section
II we present a background on the network protocols and the
software platform and tools used. In Section III we present the
study methodology including the description of the simulation
1 www.contiki-os.org
TABLE I.
Layer
Application
Transport
Network
Adaptation
Data link
Radio Duty Cycling
Physical
C OMMUNICATION PROTOCOL STACK .
Protocol
CoAP
UDP
IPv6 / RPL
6lowpan
IEEE 802.15.4 MAC (CSMA)
ContikiMAC
IEEE 802.15.4 PHY
Standard
IETF RFC 7252
IETF RFC 768
IETF RFC 6550
IETF RFC 6282
IEEE 802.15.4
IEEE 802.15.4
environment and the data analysis methodology. Section IV
shows the results of the simulations, including a deep analysis
on one node’s average power consumption and its variation
with the change of some parameters. Finally Section V concludes the paper and states future work.
II.
BACKGROUND
In this section we review the main characteristics of the
protocols, the software platform and tools used.
A. Network protocols
IEEE 802.15.4 [5] is the de facto standard for low-rate
wireless personal area networks (LR-WPANs), which specifies
the physical and media access control layers. The Medium
Access Control (MAC) layer design is crucial in wireless
sensor networks, since it directly controls the radio transceiver,
usually the most power-consuming component of the system.
Consequently, in the first years of WSN research, the focus
was on MAC protocols [6], [7], with the idea of keeping
the radio off as much time as possible. The ContikiMAC
protocol [8], [9] has integrated the main ideas of previous duty
cycling protocols, such as low power listening and phase-lock
optimization. In the following years, research headed towards
the design of a complete IPv6-based network architecture for
wireless sensor networks [10], proposing an adaptation layer
to enable the transport of IPv6 packets over IEEE 802.15.4
networks. Finally, the CoAP application-layer protocol was
designed specially for WSN. It provides a REST-like interface
[11] on top of the UDP transport layer, but with a lower cost
than HTTP-based REST interfaces.
Table I shows the analyzed network protocols’ stack in this
work, which is among the most widely used in WSN. Almost
every layer is standardized by the Internet Engineering Task
Force (IETF), except the physical and media access control
layers which are standardized by IEEE.
1) CoAP: The Constrained Application Protocol (CoAP) is
a RESTful protocol for use with constrained hardware such as
WSN nodes. The REST model works with server nodes that
make certain resources available under a URL, and client nodes
that access these resources using methods such as GET, PUT,
POST, etc. In this work we will use the OBSERVE mechanism,
which allows client nodes to retrieve a resource value from a
server (GET) and keep it updated over a period of time. From
hereon this period will be referred as observe period.
2) RPL: The IPv6 Routing Protocol for LLNs, or simply
RPL, is a proactive routing protocol based on a tree-oriented
strategy, or more specifically a Destination-Oriented Directed
Acyclic Graph (DODAG). The distance metric is usually based
on some link quality indicator, and one of the most widely
used is the Minimum Rank with Hysteresis Objective Function
(MRHOF). The tree topology is automatically built by the
nodes exchanging ICMPv6 control packets to find and propagate the routes in the network. RPL enables three operation
modes in which different kinds of traffic is supported. This
work is restricted to the Storing Mode of Operation with no
multicast support, in which each network node stores both the
table entries to route packets to all the nodes downwards from
the tree and the default route to the root.
There are three types of RPL messages: DODAG Information Solicitation (DIS), DODAG Information Object (DIO)
and Destination Advertisement Object (DAO). The former two
messages are sent as link-layer broadcast. DIS messages are
requests to join a DODAG, whereas DIO messages contain
information about the DODAG the sender node belongs to.
Essentially, the DIO messages inform the distance to the root
node using the DODAG metric. Based on this information
received from its neighbors, a node selects its prefered parent
for sending messages to the root. The Trickle algorithm [12] is
used to control when DIO messages are sent, and it is based
on a timer whose duration is doubled each time it is fired,
sending fewer messages per unit of time when the network is
stable.
3) 6LoWPAN: 6LowPAN is an adaptation layer protocol
that allows the transport of IPv6 packets over 802.15.4 links.
It is in charge of the compression of IPv6 and the upper
layer headers and of the fragmentation and reassembly of IPv6
packets. When we refer to 6LoWPAN packets in section IV,
we are referring to the packets of the application layer that
have been fragmented using this adaptation layer in order to
be transmitted.
4) CSMA: The IEEE 802.15.4 standard defines the use
of Carrier Sense Multiple Access with Collision Avoidance
(CSMA/CA) in the MAC layer. This mechanism does carrier
sensing before transmitting every packet to check whether
the channel is idle or not. Once the channel is detected
idle for transmission, the packet is transmitted. In Contiki N
carrier senses are performed before transmitting, where N is
a configurable integer with a default value of 6.
5) ContikiMAC: ContikiMAC is a radio duty cycling
(RDC) protocol based on the Low Power Listening (LPL)
mechanism, that uses periodic wake-ups (Clear Channel Assessment or CCA) to listen for packet transmissions from
neighbor nodes. This enables the radio transceiver to achieve
duty cycles below 1% [13], making an important reduction
on the energy consumption in comparison with the systems
that do not manage the RDC layer. LPL determines a baseline
energy consumption that can be calculated from the CCR. If
during a periodic wake-up a transmission is detected, the radio
transceiver is kept on in order to receive the packet. After the
packet is successfully received, a link-layer acknowledgement
is sent.
For the transmission of packets, the phase lock mechanism
is used. Every time a node sends a packet to a neighbor,
it records the time at which the neighbor replied with an
acknowledgment, learning its phase. Using the phase information, the sender does not start to send its strobes until the
neighbor is awake. If the sender misses the awake period of
the neighbor, the sender will simply continue to send strobes
for an entire period. Broadcast transmissions do not receive
Fig. 1. Sensor node states estimated by Energest. Boxes in gray represent
calculated times using Energest estimations.
link layer acknowledgements and they are a bit different from
unicast transmissions. Broadcasts are intended to be received
by every neighbour node, so the sender needs to repeatedly
send the packet during the full wake-up period. The CSMA/CA
mechanism described in Section II-A4 only senses the channel
N times for the first packet of the transmission burst to avoid
collisions. From hereon the wakeup interval will be referred
as CCR.
B. Software platform and tools
1) Contiki OS: Contiki OS is an open source, event-driven
operating system oriented to WSN applications using constrained hardware. This OS manages the hardware resources
and includes different libraries such as network stacks and
a file system. Contiki OS’s scheduler manages sleep modes
powering down the microprocessor when there is neither
processing needed nor events scheduled in the event queue.
2) Energest Module: Energest is a software-based online energy estimation mechanism implemented as a Contiki
module, that measures the accumulated time the sensor node
is in different states such as CPU, LPM, IRQ, Tx and Rx. The
mechanism runs directly on the sensor nodes and provides
real-time estimates of the current energy consumption. The
different states are represented in Figure 1. LPM state is
activated when the sensor node goes to LPM but is not
deactivated in the interrupt requests handling, so the actual
time the sensor node is in LPM is LP M − IRQ. CPU
state is activated whenever the node is active, including radio
transmissions, so the real time the CPU is active without using
the radio transceiver is CP U − T x − Rx.
Taking this into account, we can calculate the average
power as stated on equation 1.
tstate
state
Pavg
= DCstate × Istate × V =
× Istate × V (1)
ttotal
where Istate is current consumption at a certain state and V
is the operating voltage.
The current consumptions of each state were taken from
the MSP430F5438 [14] and CC2420 [15] datasheets, and the
operating voltage was 3V . Note that for Tx and Rx states,
CPU current consumption is taken into account.
3) Cooja Network Simulator: For the analysis of the network and the hardware emulation we used the Cooja Network
Simulator [16]. Cooja is a Java-based network simulator capable of emulating several platforms with radio communication.
III.
M ETHODOLOGY
A. Simulation environment
We chose to emulate in the Cooja Network Simulator the
EXP-5438 platform, which consists of an MSP430F5438 [14]
Fig. 2.
Network topology
MCU and a CC2420 [15] radio. We selected this platform
because it is very similar to the well-known Tmote-Sky [17]
that is used as reference in many research works, but it has
more ROM memory that has allowed us to compile more
complex applications. As radio medium, we used Multi-path
Ray-tracer Medium (MRM) [18] and configured its parameters
in order to obtain a link quality of 95% between neighbor
nodes. This value was chosen in order to obtain a good
link [19]. The network topology is shown in Figure 2 and
was comprised by 10 EXP-5438 nodes that used the network
protocols described in Section II.
The node with ID 1 is a border router and also the RPL
network sink. This node is the only one that has the RDC
protocol disabled and it observes a CoAP resource on each
of the other nine nodes. Nodes with IDs 2 to 10 are CoAP
servers with observable resources. Between simulations, CCR
and the observe period of the resources was changed in order
to analyze their influence on the average power consumption.
In order to select the simulation time, a long-term simulation was executed (24 hours) and the variation of the duty
cycle of each node state was analyzed. As it is shown on
Figure 3, after around 4 hours of simulation the duty cycle of
each state remains steady, so this length of time is enough for
our analysis. Figure 3 shows the first 12 hours of simulation
because the rest did not contain relevant information.
During the simulation, each network packet was captured
and recorded using the Cooja Radio Logger for later analysis.
B. Data analysis
The main objective of the data analysis was to break down
the average power consumption of a node according to the
different network protocols. In order to accomplish this task,
we analyzed the node with ID 2 as it is representative of
our network. We chose to study the variations in average
power consumption in order to detach from the simulation
time variable.
To break down the average power consumption of the
radio in Tx and Rx states according to the different network
protocols, we had to create an empirical model to relate the
where Nbroadcast is the number of broadcast transmissions
and Nunicast is the number of unicast transmissions.
This analysis was repeated for different CCRs (2, 4 and
8Hz) and different observe periods of the CoAP resources
(6 and 60s). We then validated our model comparing the
estimated total time for Tx and Rx as described before with
the Energest Tx and Rx reported times. An error minor than
1, 5% was achieved in every case.
IV.
R ESULTS
In this section we present the results of the simulations and
data analysis described in Section III.
Fig. 3.
Variation of different state’s duty cycle in time
A. Average power consumption breakdown
As a case of study, we made a deep analysis of the average
power consumption with a configuration of CCR at 4Hz and
an observe period of 60s. In Section IV-B we will compare
these results with the ones obtained with other configurations.
Table II shows the results of the Energest module, where it
can be observed that the radio Rx state has the higher average
power consumption, followed by the LPM, CPU, radio Tx and
IRQ states.
TABLE II.
Fig. 4.
E NERGEST RESULTS WITH CCR AT 4Hz AND AN OBSERVE
PERIOD OF 60s
N ode state
CPU
TX
RX
LPM
IRQ
Time spent sending or receiving a packet as a function of the size
size of the packets (information we obtain from the packet)
with the radio time spent transmitting and receiving them.
For this purpose we analyzed the Cooja timeline and recorded
the time spent transmitting and receiving several packets of
different sizes. We found out that they have a linear relation
with an offset as shown in Figure 4. The curve slope is 32µs/B
which is in correspondence with the nominal IEEE 802.15.4
data rate [5]. After obtaining this model, we proceeded to
estimate the elapsed time to transmit and receive every packet
and then calculated the average power consumption related to
each network protocol as shown on Section IV. To analyze the
impact of broadcast traffic, we also broke down the ICMPv6
average power consumption according to the different RPL
control packets (DIO, DIS and DAO). Besides the packet
times, we also had to take into account the time the radio
is in Rx state due to the LPL and CSMA as described on
section II. These times were calculated following [8].
tsim
× 2 × tr
2 × tr + tc + tCCA
(2)
where tsim is the simulation time, tCCA is the inverse of the
CCR, 2 × tr is the time elapsed performing two consecutive
CCAs and tc is the time between consecutive CCAs.
The CSMA time is calculated as:
tCSM A = 6 × tr × (Nunicast + Nbroadcast )
(3)
Pavg (µW )
114.51
64.35
227.4
163.5
75.6
Using the methodology described in Section III-B the
average power of radio Tx and RX states given by Energest
can be decomposed in the different network protocols as stated
on Table III. The application layer traffic is the one under
the CoAP row summed up to the one under 6LowPAN row.
This is because when CoAP messages are fragmented, the first
fragment is identified as a 6LoWPAN packet. When the final
fragment arrives, the packet is reassembled and recognized as
CoAP. The only fragmented packets in our network are CoAP
messages, as RPL control packets do not have the size to be
fragmented.
TABLE III.
AVERAGE POWER CONSUMPTION BREAKDOWN WITH CCR
AT 4Hz AND AN OBSERVE PERIOD OF 60s
Tx
The LPL time is calculated as:
tLP L = N × tON =
DC(%)
0.61
0.11
0.35
98.53
0.40
Rx
P rotocol
CoAP
ICMPv6
6LoWPAN
IEEE 802.15.4
CoAP
ICMPv6
6LoWPAN
LPL
CSMA
IEEE 802.15.4
DC(%)
0.0097
0.0945
0.00554
0.00153
0.0058
0.0092
0.0011
0.30
0.026
0.0023
Pavg (µW )
5.70
55.29
3.24
0.90
3.81
6.03
0.75
195.18
16.95
1.50
It can be observed that the CoAP application packets
have a very low weighting over the total average power
(around 15% for Tx and negligible for Rx), so increasing the
CoAP resources observe period would not have great impact
on reducing the average power consumption. It can also be
observed that in this case the CoAP Tx average power is 50%
higher than Rx since the node with ID 2 transmits it’s own
CoAP messages and also routes the packets from node with
ID 7. The radio Rx average current is dominated by the LPL,
and packet traffic weighs less than 5%. Since the ICMPv6
RPL messages represent over 85% of the Tx average power
consumption, it is worth analysing them carefully. Table IV
shows the consumption breakdown of ICMPv6 RPL messages.
DIO and DIS are broadcast packets and they represent over
95% of the total RPL packets.
TABLE IV.
DC(%)
0.0911%
0.0032%
0.0002%
Pavg (µW )
53.30
1.87
0.11
B. Comparison of results changing CCR and observe period
When analysing the average power consumption calculated
from the results of the Energest module varying the CCR and
observe period, we obtain the results shown on table V. It can
be observed that the total average power consumption reduces
around 30% when lowering CCR from 8Hz to 2Hz. Also if
we reduce the observe period from 60s to 6s, the total average
power consumption increases around 60%.
If we consider for example a platform using a Power
Amplifier/Low Noise Amplifier (PA/LNA) like the TI CC2592
[20], the transmit power is seven times higher, so the increase
in the Tx average power due to the decrease of the CCR could
make the total average consumption to increase.
CPU
TX
RX
LPM
IRQ
Total
Tx
Rx
AVERAGE POWER CONSUMPTION BREAKDOWN WITH CCR
AT 2Hz AND AN OBSERVE PERIOD OF 60s
P rotocol
CoAP
ICMPv6
6LoWPAN
IEEE 802.15.4
CoAP
ICMPv6
6LoWPAN
LPL
CSMA
IEEE 802.15.4
DC(%)
0.010
0.17
0.0065
0.0015
0.0059
0.0049
0.0011
0.15
0.027
0.0019
Pavg (µW )
5.94
99.75
3.84
0.87
3.84
3.24
0.72
97.86
17.46
1.23
Dif f (%)
4.55
80.45
18.78
-2.77
1.56
-40.35
-4.08
-49.87
3.05
-17.71
RPL T X AVERAGE POWER BREAKDOWN FROM TABLE III
Control P acket
DIO
DAO
DIS
TABLE V.
TABLE VI.
E NERGEST RESULTS VARYING CCR AND OBSERVE PERIOD
8Hz, 60s
Pavg (µW )
51,61
38,69
418,25
160,62
146,47
815,64
4Hz, 60s
Pavg (µW )
114,51
64,34
227,41
163,50
75,59
645,34
2Hz, 60s
Pavg (µW )
141,29
109,85
130,61
161,28
40,23
583,26
4Hz, 6s
Pavg (µW )
237,42
166,72
388,46
159,34
74,63
1026,57
According to the protocols described in Section II and the
presented model, we expect to observe the following effects
on the average power consumption when the CCR is reduced
from 4Hz to 2Hz: a) Tx times increase as broadcast times
double, b) Rx times decrease as LPL times reduce to half, c)
CPU times increase according to the increase of processing
times due to the handling of more radio packets and d) IRQ
time reduces due to the decrease of radio interruptions for the
CCA.
These effects are reflected on the simulation results obtained with a configuration of CCR at 2Hz and an observe
period of 60s as shown on Table VI. Dif f column shows the
variation with the simulation at 4Hz and an observe period
of 60s. It is interesting to remark that in Tx the ICMPv6
packets traffic (mostly broadcast) increases by 80%, while
CoAP, 6LoWPAN and 802.15.4 packets traffic change less than
20%.
In order to examine how the observe period influences
the average power consumption, we performed a simulation
maintaining the CCR at 4Hz but decreasing 10 times the
observe period to 6s. We expected to see several effects on
the average power consumption such as: a) LPL times remain
equal, b) the RPL traffic increases, c) CoAP, 6LoWPAN and
802.15.4 packet traffic increases 10 times, d) CSMA times
increase 10 times, e) CPU times increase according to the
increase of processing times due to the handling of more
radio packets and f) IRQ times remain equal because most
of the interruptions belong to the radio’s LPL which remains
constant.
TABLE VII.
AVERAGE POWER CONSUMPTION BREAKDOWN WITH
CCR AT 4Hz AND AN OBSERVE PERIOD OF 6s
Tx
Rx
P rotocol
CoAP
ICMPv6
6LoWPAN
IEEE 802.15.4
CoAP
ICMPv6
6LoWPAN
LPL
CSMA
IEEE 802.15.4
DC(%)
0.088
0.14
0.057
0.0080
0.062
0.014
0.0010
0.30
0.19
0.019
Pavg (µW )
51.69
82.5
33.12
4.74
40.68
9.36
0.93
195.18
126.03
12.66
Dif f (%)
807.68
49.23
921.02
429.42
969.28
55.19
24.85
-0.01
644.11
737.26
The obtained simulation results are shown on Table VII,
and even though the effects are not exactly identical to the
ones described in the previous paragraph, they are very similar.
Dif f column also shows the variation with the simulation at
4Hz and an observe period of 60s. Variations in CPU and
IRQ average power consumption can be observed on table V.
C. Results analysis
The results confirm that using protocols that achieve very
low radio duty cycles is a great advantage as the system is
in Low Power Mode over 98% of the time. Nevertheless, the
Radio’s Rx and Tx states consume a considerably high share
of the overall power. Modifying the CCR has a great impact
on the radio’s Rx and Tx states average power consumption.
Reducing the CCR decreases the radio’s Rx average power
consumption while it also increases the radio’s Tx average
power consumption. If we consider for example a platform
using a power amplifier, the Tx average power consumption
would increase considerably. In this case if we reduce the
CCR, even though the Rx average power would decrease,
the total average power consumption could increase. Also
if we consider very lossy networks, the radio’s Tx duty
cycle could increase due to the increase in RPL DIOs, so
the convenience of reducing the CCR should be carefully
analyzed. The presented approach shows an efficient way to
perform this analysis.
V.
C ONCLUSIONS AND FUTURE WORK
The average power consumption of a sensor node using the
IPv6 over LLNs was deeply analyzed, breaking it down according to the different network protocols. We observed that the Rx
state of the radio has the highest average power consumption
in almost every configuration due to the LPL mechanism.
The Tx state of the radio also has a considerable contribution
to the total average power consumption and most of the Tx
traffic is due to broadcast RPL control messages. Reducing
the CCR decreases the Rx average power consumption, while
it increases the Tx one. Convenience of modifying it should be
carefully analyzed taking into account the amount of broadcast
messages transmitted and the absolute values of the Tx and
Rx average power. Dynamically changing the CCR could be a
good solution for reducing the energy consumption in unstable
networks (e.g. due to time-varying propagation conditions or
mobile nodes).
The impact of the application report rate was also analyzed.
If an observe period of 60s is used, the average power
consumed to transmit and receive the application layer packets
is negligible compared to the average power consumed by
RPL control messages in Tx and LPL in Rx, meaning that
increasing the observe period would not impact on the total
average power consumption. If we reduce the observe period
to 6s the average power consumed to transmit and receive the
application layer packets dramatically increases and becomes
comparable to the average power consumed to transmit the
RPL control messages.
Future work should include further analysis considering:
a) different packet error rates between neighbor nodes, b)
different topologies (i.g. linear, grid, random), and c) higher
network size . When changing the network’s topology or size,
nodes at different positions should be analyzed (i.g. nodes
near to the root, leaf nodes, intermediate nodes), since the
application traffic and RPL messages (DAO) depend on the
number of nodes downwards. Finally, measurements on real
sensor nodes would give information to confirm the simulation
results.
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
ACKNOWLEDGMENTS
This work was partially funded by INIA-FPTA and a
research scholarship granted by ANII.
R EFERENCES
P. Dutta and A. Dunkels, “Operating systems and network protocols
for wireless sensor networks,” Philosophical Transactions of the
Royal Society A: Mathematical, Physical and Engineering Sciences,
vol. 370, no. 1958, pp. 68–84, Jan. 2012. [Online]. Available:
http://dx.doi.org/10.1098/rsta.2011.0330
[2] V. Raghunathan, C. Schurgers, S. Park, and M. B. Srivastava,
“Energy-aware wireless microsensor networks,” Signal Processing
Magazine, IEEE, vol. 19, no. 2, pp. 40–50, Mar. 2002. [Online].
Available: http://dx.doi.org/10.1109/79.985679
[3] M. Kovatsch, S. Duquennoy, and A. Dunkels, “A Low-Power CoAP
for Contiki,” in Mobile Adhoc and Sensor Systems (MASS), 2011
IEEE 8th International Conference on, ser. MASS ’11. Washington,
DC, USA: IEEE, Oct. 2011, pp. 855–860. [Online]. Available:
http://dx.doi.org/10.1109/mass.2011.100
[4] H. Ali, “A Performance Evaluation of RPL in Contiki,” Master’s thesis,
MSc thesis, Blekinge Institute of Technology, 2012.
[17]
[1]
[18]
[19]
[20]
“IEEE Standard for Local and metropolitan area networks–
Part 15.4: Low-Rate Wireless Personal Area Networks
(LR-WPANs),”
Tech.
Rep.,
2011.
[Online].
Available:
http://dx.doi.org/10.1109/ieeestd.2011.6012487
E. H. Callaway, “The Wireless Sensor Network MAC,” in Handbook
of Sensor Networks. John Wiley & Sons, Inc., 2005, pp. 239–276.
[Online]. Available: http://dx.doi.org/10.1002/047174414x.ch8
A. Bachir, M. Dohler, T. Watteyne, and K. K. Leung, “MAC
Essentials for Wireless Sensor Networks,” Communications Surveys &
Tutorials, IEEE, vol. 12, no. 2, pp. 222–248, 2010. [Online]. Available:
http://dx.doi.org/10.1109/surv.2010.020510.00058
A. Dunkels, “The ContikiMAC Radio Duty Cycling Protocol,” Swedish
Institute of Computer Science, Tech. Rep. T2011:13, Dec. 2011.
[Online]. Available: http://www.sics.se/˜adam/dunkels11contikimac.pdf
A. Dunkels, L. Mottola, N. Tsiftes, F. Österlind, J. Eriksson,
and N. Finne, “The announcement layer: beacon coordination
for the sensornet stack,” in Proceedings of the 8th European
conference on Wireless sensor networks, ser. EWSN’11. Berlin,
Heidelberg: Springer-Verlag, 2011, pp. 211–226. [Online]. Available:
http://portal.acm.org/citation.cfm?id=1966270
J. W. Hui and D. E. Culler, “IP is dead, long live IP for
wireless sensor networks,” in Proceedings of the 6th ACM conference
on Embedded network sensor systems, ser. SenSys ’08. New
York, NY, USA: ACM, 2008, pp. 15–28. [Online]. Available:
http://dx.doi.org/10.1145/1460412.1460415
R. T. Fielding and R. N. Taylor, “Principled Design of
the Modern Web Architecture,” ACM Trans. Internet Technol.,
vol. 2, no. 2, pp. 115–150, May 2002. [Online]. Available:
http://dx.doi.org/10.1145/514183.514185
P. Levis, N. Patel, D. Culler, and S. Shenker, “Trickle: A Self-regulating
Algorithm for Code Propagation and Maintenance in Wireless Sensor
Networks,” in Proceedings of the 1st Conference on Symposium
on Networked Systems Design and Implementation - Volume 1, ser.
NSDI’04. Berkeley, CA, USA: USENIX Association, 2004, p. 2.
[Online]. Available: http://portal.acm.org/citation.cfm?id=1251177
L. Steinfeld, M. Ritt, F. Silveira, and L. Carro, “Optimum design
of a banked memory with power management for wireless sensor
networks,” Wireless Networks, pp. 1–14, Jul. 2014. [Online]. Available:
http://dx.doi.org/10.1007/s11276-014-0763-5
“MSP430FR573x, MSP430FR572x Mixed Signal Microcontroller (Rev.
H),” http://www.ti.com/lit/ds/symlink/msp430fr5739.pdf, Texas Instruments, Sep. 2013.
2.4 GHz IEEE 802.15.4/ZigBee-Ready RF Transceiver (Rev. B).,
http://focus.ti.com/docs/prod/folders/print/cc2420.html., Texas Instruments Inc.
J. Eriksson, F. Österlind, N. Finne, N. Tsiftes, A. Dunkels,
T. Voigt, R. Sauter, and P. J. Marrón, “COOJA/MSPSim:
interoperability testing for wireless sensor networks,” in Proceedings
of the 2nd International Conference on Simulation Tools and
Techniques, ser. Simutools ’09. ICST, Brussels, Belgium, Belgium:
ICST (Institute for Computer Sciences, Social-Informatics and
Telecommunications Engineering), 2009, pp. 1–7. [Online]. Available:
http://dx.doi.org/10.4108/icst.simutools2009.5637
J. Polastre, R. Szewczyk, and D. Culler, “Telos: enabling
ultra-low power wireless research,” in Information Processing
in Sensor Networks, 2005. IPSN 2005. Fourth International
Symposium on. IEEE, Apr. 2005, pp. 364–369. [Online]. Available:
http://dx.doi.org/10.1109/ipsn.2005.1440950
C. A. Boano, K. Römer, F. Osterlind, and T. Voigt, “Demo Abstract:
Realistic Simulation of Radio Interference in COOJA,” in European
Conference on Wireless Sensor Networks EWSN 2011. Citeseer, p. 36.
K. Srinivasan, P. Dutta, A. Tavakoli, and P. Levis, “An Empirical Study
of Low-power Wireless,” ACM Trans. Sen. Netw., vol. 6, no. 2, Mar.
2010. [Online]. Available: http://dx.doi.org/10.1145/1689239.1689246
“CC2592
2.4-GHz
Range
Extender,”
http://www.ti.com/lit/ds/symlink/cc2592.pdf,
Texas
Instruments,
Feb. 2014.